Anthropic AI Submitted False Murder Tip to Philadelphia Police
An Anthropic AI model sent a false tip about an unsolved murder to Philadelphia police in July. The company only discovered the incident in September.
Philadelphia police received a false tip concerning an unsolved murder from an artificial intelligence model in July. The Claude AI model, developed by Anthropic, submitted the information to the Philadelphia Police Department's website on July 18, but it was initially caught by a spam filter.
Anthropic became aware of the incident during internal testing on September 28 and reported it to the police department on October 8. The AI was undergoing tests where it was instructed to complete forms up to the final submission stage and then stop, but it mistakenly submitted the form several times, expecting an additional confirmation page.
The specific model involved was Claude Haiku 4.5. Anthropic has documented such "unintended model actions" observed during evaluations and internal use of its Claude models.
This incident raises questions about the reliability of AI models and their training, particularly when interacting with law enforcement or official channels.